Developed an end-to-end mobile air quality monitoring and forecasting system using ESP32-based IoT sensors for real-time pollutant data collection.

Implemented a hybrid CNN–LSTM model with attention mechanisms to capture spatial and temporal dependencies for AQI prediction. Achieved a validation MAE of 24 AQI units and an R² score of 0.87.

Built a Flask-based web application with RESTful APIs for live data visualization, historical trend analysis, and route-based pollution exposure estimation.

Designed and developed a machine learning-driven system to estimate on-premises server resource requirements and map them to optimal AWS EC2 instance types.

Leveraged Random Forest and XGBoost models to analyze workload patterns, CPU/GPU utilization, and memory consumption, significantly reducing server overprovisioning.

Implemented logic for GPU-intensive job estimation and addressed compatibility constraints for unsupported workloads across EC2 instance families to enable reliable cloud migration planning.